Over the past few months, as noted earlier, with the help of research assistants I have been compiling a spreadsheet that records each occurrence of a biblical verse cited in the Bavli. The purpose of this data is not so much to ask qualitative questions (e.g., where and how does the Bavli cite a particular verse?) but to allow for quantitative analysis that might lead to new questions and avenues of investigation.
As I slowly gain more familiarity with the many extraordinary but poorly documented powers of Excel, I’ve just begun to analyze this data. Here are a couple of preliminary observations:
1. The Bavli cites somewhere in the neighborhood of 5900 discrete verses of the Hebrew Bible. The Hebrew Bible contains approximately 23,700 discrete verses. That equates to about 25% of the Bible; meaning, of course, that 75% of the Bible is never cited. It is worth noting that 3,295 verses are cited only a single time in the Bavli. I am not yet sure what to make of this – one next step is to analyze the density of citations by biblical book. Does the Bavli prefer citing from certain books, especially when the size of the book is also taken into account.
2. The seven most-cited verses (with NRSV translations, and some surrounding material added for context) are:
- Deuteronomy 24:1 (37 times): Suppose a man enters into marriage with a woman, but she does not please him because he finds something objectionable about her, and so he writes her a certificate of divorce, puts it in her hand, and sends her out of his house; she then leaves his house and goes off to become another man’s wife.
- Numbers 5:13 (29 times): If any man’s wife goes astray and is unfaithful to him, if a man has had intercourse with her but it is hidden from her husband, so that she is undetected though she has defiled herself, and there is no witness against her since she was not caught in the act; if a spirit of jealousy comes on him, and he is jealous of his wife who has defiled herself; or if a spirit of jealousy comes on him, and he is jealous of his wife, though she has not defiled herself…
- Leviticus 25:5 (24 times): You shall not reap the aftergrowth of your harvest or gather the grapes of your unpruned vine: it shall be a year of complete rest for the land.
- Numbers 30:3 (24 times): When a woman makes a vow to the Lord, or binds herself by a pledge, while within her father’s house, in her youth, and her father hears of her vow or her pledge by which she has bound herself, and says nothing to her; then all her vows shall stand, and any pledge by which she has bound herself shall stand. But if her father expresses disapproval to her at the time that he hears of it, no vow of hers, and no pledge by which she has bound herself, shall stand; and the Lord will forgive her, because her father had expressed to her his disapproval.
- Leviticus 2:2 (21 times): After taking from it a handful of the choice flour and oil, with all its frankincense, the priest shall turn this token portion into smoke on the altar, an offering by fire of pleasing odour to the Lord
- Numbers 6:5 (21 times): All the days of their nazirite vow no razor shall come upon the head; until the time is completed for which they separate themselves to the Lord, they shall be holy; they shall let the locks of the head grow long.
- Leviticus 6:3 (20 times): When any of you sin and commit a trespass against the Lord by deceiving a neighbour in a matter of a deposit or a pledge, or by robbery, or if you have defrauded a neighbour, or have found something lost and lied about it—if you swear falsely regarding any of the various things that one may do and sin thereby— when you have sinned and realize your guilt, and would restore what you took by robbery or by fraud or the deposit that was committed to you, or the lost thing that you found…
Five of these verses deal with matters of civil law; three deal with women. Why these verses in particular, though? Verses dealing with some expected topics, such as Shabbat or circumcision, are absent.
I have some ideas for at least some of the verses. Deuteronomy 24:1-4 are the basis for almost the entire legal institution of divorce – the rabbis need to keep appealing to them for authority, perhaps at a time when most Jews would have respected the Bible far more than rabbinic say-so. Similarly, the sotah (“suspected wife,” in Numbers 5), issues dealing with female vows (Numbers 30:3), and the nazirite vow (Numbers 6:5) are dealt with only in these places and all generate a large body of laws. I am not yet entirely satisfied with these explanations, and would welcome yours as well!
Michael Satlow is a professor of religious studies and Judaic studies at Brown Universityand has been a mentor and sounding-board for the New Talmud Blog from the beginning. This post was crossposted from his own blog, Then and Now.
The Talmud Blog 
I’m not sure if you did this yet, but for frequency analysis you need to install the analysis toolpack into Excel, which is not enabled by default, and that gives you access to running histograms and such which makes things much easier. For clarity a bar chart is commonly used to represent the frequency of individual events, not a line graph. Good luck with the analysis.
It may be worthwhile to investigate the number of citations per topic, rather than focusing on individual verses. A topic which is based on a few biblical verses will be called into service many times (How many verses actually deal with divorce? If it’s only one or two, then the verses dealing with it are likely to be cited several times in various contexts). On the other hand, verses which deal with a topic which is broadly rooted may each, by themselves, only be cited a small handful of times in the Talmud (Kashrut and Shabbat are each discussed in several places in the Torah. Idolatry, besides the references thereto in the Torah, is dealt with throughout large parts of Nach.)
Thus, a large number of citation of a single verse may indicate the brevity in which its subject was dealt with in the Bible, rather than its importance in the eyes of the Talmudic redactor. (Cross-posted on Then and Now.)
I would tend to agree that the more information that needs to be derived from the verse the more it will be cited. so for example off the top of my head Deut 24:1 is cited regarding the proper grounds for divorce (“ervat davar” or anything displeasing), how divorce is effected (with a “sefer kritut”) how a get is constructed (writing) how it takes effect (by delivery) how it is to be delivered (into her hand) and the casuistic analysis of deviations from these requirements (does sefer require parchment, does qriting require ink, does giving rule out taking, does hand exclude lap etc.
Sotah may be a somewhat analogous but unique case where you have an incredibly elaborate construction/imagining of the entire ritual and of the various permutations of precipitating circumstances and outcomes.
Please, Please, get a statistics graduate student to help you with your analyses. This is an extremely important area in which you have used only Freshman statistical tools.
If the data set only has two fields per entry, verse and location in the Bavli (and location by daf does not give you an accurate measure of distance), I don’t think you could do that much more with the data.
Say you wanted to attempt a sophisticated mutual information analysis or derive some other information on a larger scale- what values would you want to have? Verse, location in Bavli by daf, by perek, by absolute word distance, who cited it, generation of citer, topic of sugya, purpose of citation (halakhic derasha, aggadic derasha, etc.)
Don’t ask me ask the the hypothetical statistics graduate student. Mathematics and Statitistics are intersecting magesteria (-:
Thank you thus far for these thoughtful comments. This is just what I need, except with a bit more specificity. Let me be a bit more transparent about my data, and then explain a bit more clearly what advice I think would be helpful.
The data is very simply the scriptural index from the Artscroll Talmud volumes entered into an Excel spreadsheet. Each entry is thus entered, as a row, along the following columns: (1) Book of Bible; (2) Chapter; (3) Verse; (4) Tractate; (5) page within tractate. That’s it: some 13,200 rows.
To say that my statistical skills are at the freshman level I take as a compliment. The first, and pretty much only, thing that we did was to create a concatenation of biblical book, chapter, and verse, and then count frequency and distinct entries (actually, I used a pivot table for this, which is an amazing tool). This particular visualization was a bit of a lark, just meant to serve as a graphic for the post.
What I think would be most useful is a particular set of questions that can be posed of the data that I have, with specific ideas for how to query the data and/or answer the questions. Here, the collaborative approach is most promising – some may be better equipped to ask the questions, and others have ideas and know-how about how to answer them. Ultimately, perhaps this could even result in a fuller collaboration, with others working with the data I’ve assembled.
This project begs for a open, structured, digital version of the text that fully expresses the component layers, subject matter, intra- and intertextual relationships, etc. of the Talmud http://goshabrarian.tumblr.com/post/18733129606/asking-quantitative-questions-of-the-talmud-bavli
Agreed, Joshua, although the Perseus model, I think, is not really ideal. So what would be involved to create a TEI tagged text of the Bavli with the data you would like to see included? Do you have a proposal in the works?
You’re right about the specific implementation of Perseus not being perfect, mainly because the Talmud and related materials are so much more complex. I also don’t think that TEI is quite right either, since it too isn’t able to fully cope with all of the possibilities of the text. I’ve experimented a bit with creating an XML schema that addresses the layers of the text, supports links to commentaries, biographical dictionaries, etc., but the project is beyond both my technical and Talmud skills. I’ve got some thoughts on the matter, though, and know a couple of other people who are interested in this problem and might be worth talking to.
I am trying to figure out the direction that this research might take. Why does the Bavli cite verses to begin with? Are you planning on distinguishing between halachic and aggadic citations? If so, then with the subset of citations for halachic purposes, it might also be instructive to see where a sugya manages to not cite a verse. In other words, the Bavli (and certainly the Mishna) often seems to assume that the verses are known in terms of laying out the basic topic, and thus verses are not generally cited at the beginning of a topic – since you should have read the Torah on your own.
Wondering what your plans are with this potentially interesting data.
An intriguing enterprise! I have a two more question to add to those suggested by Aaron:
How many of these quotations are “inherited” Tannaitic/Palestinian drashot, and how many are unique to the Bavli? This information is crucial: We are forced to ask no only about the place of the Bible in the Bavli or the Bavli’s attitude towards scripture, but the Bavli’s relationship with interpreted scripture, as mediated by Midrashic traditions. (I’m not sure we need to differentiate between Drashot which appear in Baraitot and “inherited” Drashot that appear in the Ammoraic/Stammaitic strata for the sake of this particular question).
But indeed — the differentiation between Amoraic/Stammaitic materials is also crucial if we wish to historicize the question of scripture in the Bavli, and not treat the Bavli as a unified work. Can we find shifts in the relationship with Scripture in the different Amoraic periods and in the work of the Stam?
Crosscommented from Satlow’s blog:
What are your specifications for including a verse? When the Talmud quotes one word and expounds on it and then quotes another word in the same verse immediately afterwards is it counted as twice or once? Furthermore, when quoting Braithot, the Talmud will quote a verse multiple times to expound different parts of it, are these counted separately or once? When the Talmud mentions various explanations of a verse, each time it offers an explanation it will quote the verse again, is this also counted as once or multiple instances? Also, the Talmud frequently will refer to a specific rule by using the exact wording of the Scripture, is this counted as an instance of quoting the verse? If so, then each time the Talmud would refer to such rules, that would be another citation of the verse. I would not rely on Artscroll because their specifications for inclusion is probably arbitrary and depends on the whim of each volume’s individual editor.
Many of the questions raised are, of course, beyond the scope of the collected data. They do point in the direction advocated by Joshua Parker: the possibilities opened by richly-tagged editions of rabbinic literature. These tags would go well-beyond those in the Bar-Ilan corpus, including the kind of strata and other information found in Talmud Ha-Iggud.
Creating and implementing such a tagging schema strikes me as an extensive undertaking that would require a good deal of collaborative work, new research, and money. It is, though, certainly possible. Anyone want to take the lead in developing a proposal for the NEH or Israel Academy?
Is it possible to compile the as data the period when the quote is cited to do a time series?
First, have you seen the scriptural index in the Ketuvim edition of the Bavli? that has been most useful to me in my research in the past. I dont know about the Artscroll edition serving as anything accurate in terms of indexing, but I can say that Neusner’s indexing is not to be trusted. His index to the Mishnah, for example, includes verses that he uses to explain or expound upon in the Mishnah text itself. Further, verses are cited in terms of page in the edition (much like Soncino or Ketuvim eds do, unfortunately), rather than by tractate and mishnah. The same can be applied to his various editions of translations of the Bavli. In addition, your dataset must take into account the added text, and the censored and missing text, that has crept into or out of the Bavli. The set should also only include the Bavli; not the mishnah, nor later additions. Finally, have you consulted, and compiled a list, of articles from the literature of the past 150 years, that discuss Scriptural citations? — there are many.
Pingback: How Open is ‘Open’? | The Talmud Blog
Reblogged this on Talmidimblogging.
https://en.wikipedia.org/wiki/Zipf%27s_law